Topic: PhoenixMiner 6.2c: fastest Ethereum/Ethash miner with lowest devfee (Win/Linux) - page 319. (Read 784965 times)

I have been playing around with these for most part of the day and started off by trying:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -mode 1 -clKernel 3 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096
But I was getting between 30% to 50% incorrect shares.
I've tried slightly adjusting the -gt between 15 to 150 (even 0 or leaving it out completely) and it seemed to help slightly at ~115 but I was still experiencing a lot of incorrect shares.
Now... I've tried experimenting with -clKernel 3 but it always said that it was ignored.

I've tried:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -clKernel 3 -clNew 1 -openclLocalWork 64
Which seemed a little better but I was still getting between 5% incorrect shares (about 1 out of every 15-20) but I was only yielding ~56MH/s. Not too bad...

So then I remove the -openclLocalWork...
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -clKernel 1 -clNew 1

I let this run for exactly 5 hours and 19 minutes.
Three XFX 5700 XT cards.
Accepted 318, Rejected 0, HW Errors 0
Averaging 52 MH/s for 156 MH/s combined and 277w combined (92w - 94w each)


You have to admit that these are very solid, stable and promising. Results of 52MH/s for 92w is 0.56MH/s per watt seems pretty efficient. Ideally I'd love to have a 500+MH/s Ethereum mining rig on a 1,000w PSU and I believe that it's achievable with these cards.

Yeah that didn't help at all with my incorrect shares. I removed the -gt parameter so it does it automatically. I'm going back to -clKernel 1 for the moment. Perhaps that will help.
And it looks like I can remove the -acm since Compute Mode isn't needed for these cards.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -clKernel 1 -clNew 1 -openclLocalWork 64
Reduces my hashrate slightly and I'm 3 for 3 so far. It's been running for a couple of minutes and no incorrect shares.

With -clKernel 3 and new founded -gt parameter (it differs from -clKernel 1) you can gain about 0,7-1MH/s additional speed...
And I recommend to find -gt parameter for each card individually, not using the same for all cards. Best of it can be differ even on two equal cards. Each changing of cclock or mclock leads to new best -gt parameter, just check it...

How can I troubleshoot incorrect shares?

I'm mining ETH with RX 5700 XT
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

I'm averaging 60MH/s but I'm getting a lot of incorrect shares. (11 out of 26)

Windows 10 with 19.9.2 driver

Wow... Someone posted on Claymore thread that following config is getting 60 MHs with Phoenix 4.6c, I have it running for 10 min and it's true. I'm getting 59.7 at 104-110 Watt. I only hope it stays stable without lockup.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

Same here. I tried every possible combination of -tt -cvddc and -mclock for RX5700 but one card would go down to 0 Mhs followed by the second card after some time. Claymore is a bit slower but it's stable with  -cvddc 800 -cclock 1300 -mclock +910 -powlim -25,-25 -tt 58,58 settings (hashing at 50.2 - 50.5 Mhs).
I'm not using 3rd party hardware control but I also have 2 vii's and 2 1060's in the same rig which are working well with phoenix.

I tried both ways, and still would get hashrate drops to 0 mhs and a miner restart... After sometime, it would just lock the rig up. I was mining Ubiq, so there might be some issues there.... However, I did switch to Nicehash/Claymore for a test and haven't had any miner restarts/rig locks in 18 hours... I'm going to come back to PM after 24 hours and connect to NH and see how it performs.

   Currently there can be only one config.txt file (or command-line options). We will probably add ability to use both config.txt file and command-line options in the next release, as it can be useful in some cases.



  It is important to avoid using both the PhoenixMiner hardware control options and third party programs (e.g. Afterburner) at the same time. While we use the AMD provided API to do most of the hardware control and monitoring, most of these programs use lower-level undocumented access and doing both at the same time can lead to driver lock ups. Note that Afterburner must be completely closed, not only minimized to tray in order to stop interacting with the GPUs. Or, alternatively, you should use only Afterburner (or other third party program) and none of PhoenixMiner hardware control options.

It went watch dog and rebooted an hour in.

Been having the same exact problem.... Most of the time, my mining watchdog wouldn't catch the reboot of the miner since it does it fairly quickly. But, I noticed a ton of log build up... After opening several, I saw the same thing (I already knew it was going to 0 mhs and locking up the rig, but I didn't know it was doing it 20 times a day)

I went ahead and trying out the AB way...

Lower -cvddc to 750. If it continuous start lowering -tt

Please support any ProgPOW such as SERO, ZANO, EPIC 

Got my 5700xt up and running last night.

Windows 1903, Driver 19.9.1 -- Running, -cclock 1300 -cvddc 800 -mclock 900 -hstats 2 - Getting roughly 52 mhs.

Seems like after some hours, (3-8 hours) threads stop responding and drops to 0 mhs.

hi,

if i use at least one command line parameter, it completely ignores the config.txt

so i cant use both options together? i have to choose either config or command line ?

Confirm. 4.6c supports 19.9.2 AMD driver version.

AMD reports GPU-Only powerdraw, means you have to add additional watts(PCIe power consumption). Some GPUs(won't get into details) also report less powerdraw than actual, its not a bug, its VRM specific.

I estimate wall power as 35-40% higher than reported GPU power in software (such as HWInfo, GPU-z, ect.). That accounts for the extra system draw, and it's been pretty close in my experience. It's actually even high in some cases with multiple gpus, and 30% over reported can be accurate.

is there a direct connection between card power draw reported and real "off the wall" power draw ?

Have updated today to 19.9.2 version. Same hashrate.